Content sharing applications have been one of the most widely used and popular applications over the Internet. At the same time, the use of wireless communication devices has become pervasive, and is rapidly overtaking the use of traditional wired devices. As a result, much content is rendered on mobile devices. To provide a satisfactory user experience, many service providers attempt to recommend relevant content to a user based on past purchases and renderings by the user or past purchases and renderings by known contacts of the user. Content is usually classified and relevance determined for recommendations based on metadata that names the content and describes the content, such as by artist name, performance date, and genre tags. However, values for these metadata parameters are often too specific or too coarse or too difficult to translate effectively into enough different languages to provide nuanced recommendations that are more helpful for the user and service providers to administer.